{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 16,
  "concurrency": 16,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-b12x-nospec",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T19:04:23Z",
  "gpu": "dual-rtx-pro-6000-tp2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 639,
    "content_chunks_s": null,
    "decode_tok_s_p50": 3.4702313418761683,
    "decode_tok_s_p95": 24.5333685160151,
    "e2e_p50_ms": 20628.623299999163,
    "e2e_p95_ms": 21218.045300000085,
    "effective_prefill_tok_s_p50": 356.5020466017027,
    "effective_prefill_tok_s_p95": 1785.311482113694,
    "generation_p50_ms": 12381.745900001988,
    "generation_p95_ms": 18358.67960000178,
    "mean_inter_token_latency_ms_p50": 275.14990888893306,
    "mean_inter_token_latency_ms_p95": 556.969841379308,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 715,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 16,
    "prompt_tokens": 47488,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 44,
    "throughput_tok_s": 33.689205367012086,
    "time_to_first_output_p50_ms": 7715.098800003034,
    "time_to_first_output_p95_ms": 19195.00490000064,
    "ttft_p50_ms": 10770.645500000683,
    "ttft_p95_ms": 19404.483099999197,
    "ttft_samples": 16
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-nospec",
  "request_timings": [
    {
      "content_chunks": 48,
      "decode_tok_s": 2.8869178587328723,
      "e2e_ms": 20628.623299999163,
      "effective_prefill_tok_s": 1307.5214156207576,
      "generation_ms": 18358.67960000178,
      "mean_inter_token_latency_ms": 346.39018113210903,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 0,
      "time_to_first_output_ms": 2269.943699997384,
      "ttft_ms": 5289.356999997835,
      "visible_generation_ms": 15339.266300001327
    },
    {
      "content_chunks": 38,
      "decode_tok_s": 2.135446556644647,
      "e2e_ms": 19925.61399999977,
      "effective_prefill_tok_s": 1785.311482113694,
      "generation_ms": 18263.158999998268,
      "mean_inter_token_latency_ms": 468.2861282050838,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 1662.4550000015006,
      "ttft_ms": 1662.4561000026006,
      "visible_generation_ms": 18263.157899997168
    },
    {
      "content_chunks": 28,
      "decode_tok_s": 1.79542934950221,
      "e2e_ms": 19626.137200000812,
      "effective_prefill_tok_s": 854.3436726378557,
      "generation_ms": 16152.125399999932,
      "mean_inter_token_latency_ms": 556.969841379308,
      "output_token_source": "usage",
      "output_tokens": 30,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 3474.0118000008806,
      "ttft_ms": 3474.0124000018113,
      "visible_generation_ms": 16152.124799999001
    },
    {
      "content_chunks": 38,
      "decode_tok_s": 2.5227963761776286,
      "e2e_ms": 20146.663200001058,
      "effective_prefill_tok_s": 633.1561520078761,
      "generation_ms": 15459.036000000197,
      "mean_inter_token_latency_ms": 396.3855384615435,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 4687.627200000861,
      "ttft_ms": 4687.627800001792,
      "visible_generation_ms": 15459.035399999266
    },
    {
      "content_chunks": 39,
      "decode_tok_s": 2.7994113067060313,
      "e2e_ms": 20405.079800002568,
      "effective_prefill_tok_s": 633.1747516447657,
      "generation_ms": 15717.590300002485,
      "mean_inter_token_latency_ms": 357.21796136369284,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 4,
      "time_to_first_output_ms": 4687.489500000083,
      "ttft_ms": 7715.753500000574,
      "visible_generation_ms": 12689.326300001994
    },
    {
      "content_chunks": 39,
      "decode_tok_s": 3.0135608113318364,
      "e2e_ms": 20491.72859999817,
      "effective_prefill_tok_s": 503.81417717157996,
      "generation_ms": 14600.667700000486,
      "mean_inter_token_latency_ms": 331.83335681819284,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 5,
      "time_to_first_output_ms": 5891.060899997683,
      "ttft_ms": 8934.585899998638,
      "visible_generation_ms": 11557.142699999531
    },
    {
      "content_chunks": 49,
      "decode_tok_s": 4.071751315888464,
      "e2e_ms": 20977.205400002276,
      "effective_prefill_tok_s": 384.7001933402114,
      "generation_ms": 13262.106599999242,
      "mean_inter_token_latency_ms": 245.59456666665264,
      "output_token_source": "usage",
      "output_tokens": 55,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 6,
      "time_to_first_output_ms": 7715.098800003034,
      "ttft_ms": 10770.645500000683,
      "visible_generation_ms": 10206.559900001594
    },
    {
      "content_chunks": 40,
      "decode_tok_s": 3.0840218819626455,
      "e2e_ms": 20404.216899998573,
      "effective_prefill_tok_s": 417.4467773644162,
      "generation_ms": 13294.32850000012,
      "mean_inter_token_latency_ms": 324.2519146341493,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 7109.888399998454,
      "ttft_ms": 7109.8886999971,
      "visible_generation_ms": 13294.328200001473
    },
    {
      "content_chunks": 33,
      "decode_tok_s": 3.4702313418761683,
      "e2e_ms": 20491.09150000004,
      "effective_prefill_tok_s": 311.0846037029357,
      "generation_ms": 10950.278600001184,
      "mean_inter_token_latency_ms": 288.16522631582063,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 8,
      "time_to_first_output_ms": 9540.812899998855,
      "ttft_ms": 12605.638599998201,
      "visible_generation_ms": 7885.452900001837
    },
    {
      "content_chunks": 40,
      "decode_tok_s": 3.63438245005438,
      "e2e_ms": 20707.083799999964,
      "effective_prefill_tok_s": 356.5020466017027,
      "generation_ms": 12381.745900001988,
      "mean_inter_token_latency_ms": 275.14990888893306,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 9,
      "time_to_first_output_ms": 8325.337899997976,
      "ttft_ms": 11373.24240000089,
      "visible_generation_ms": 9333.841399999073
    },
    {
      "content_chunks": 40,
      "decode_tok_s": 4.220240272346428,
      "e2e_ms": 20814.35070000225,
      "effective_prefill_tok_s": 292.3720055104196,
      "generation_ms": 10662.899999999354,
      "mean_inter_token_latency_ms": 236.95333333331897,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 10,
      "time_to_first_output_ms": 10151.450700002897,
      "ttft_ms": 13210.466900000029,
      "visible_generation_ms": 7603.883800002222
    },
    {
      "content_chunks": 40,
      "decode_tok_s": 4.4643965712750235,
      "e2e_ms": 20848.962900003244,
      "effective_prefill_tok_s": 275.60044962241955,
      "generation_ms": 10079.74970000214,
      "mean_inter_token_latency_ms": 223.9944377778253,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 11,
      "time_to_first_output_ms": 10769.213200001104,
      "ttft_ms": 13822.256000003108,
      "visible_generation_ms": 7026.706900000136
    },
    {
      "content_chunks": 37,
      "decode_tok_s": 4.71131965340426,
      "e2e_ms": 20670.040100001643,
      "effective_prefill_tok_s": 235.47408525973495,
      "generation_ms": 8065.680700001394,
      "mean_inter_token_latency_ms": 212.2547552631946,
      "output_token_source": "usage",
      "output_tokens": 39,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 12604.359400000249,
      "ttft_ms": 12604.359600001771,
      "visible_generation_ms": 8065.680499999871
    },
    {
      "content_chunks": 39,
      "decode_tok_s": 6.150522463121268,
      "e2e_ms": 20974.83899999861,
      "effective_prefill_tok_s": 214.7460622026112,
      "generation_ms": 7153.863799998362,
      "mean_inter_token_latency_ms": 162.58781363632642,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 13,
      "time_to_first_output_ms": 13820.975200000248,
      "ttft_ms": 19365.598699998372,
      "visible_generation_ms": 1609.2403000002378
    },
    {
      "content_chunks": 51,
      "decode_tok_s": 6.500925720214673,
      "e2e_ms": 21218.045300000085,
      "effective_prefill_tok_s": 235.4829016047387,
      "generation_ms": 8614.157799998793,
      "mean_inter_token_latency_ms": 153.82424642854988,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 14,
      "time_to_first_output_ms": 12603.887500001292,
      "ttft_ms": 19279.735600001004,
      "visible_generation_ms": 1938.3096999990812
    },
    {
      "content_chunks": 40,
      "decode_tok_s": 24.5333685160151,
      "e2e_ms": 21029.24139999959,
      "effective_prefill_tok_s": 154.62356042430085,
      "generation_ms": 1834.2364999989513,
      "mean_inter_token_latency_ms": 40.760811111087804,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 4,
      "request_index": 15,
      "time_to_first_output_ms": 19195.00490000064,
      "ttft_ms": 19404.483099999197,
      "visible_generation_ms": 1624.7583000003942
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260829T190402Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": "low",
    "shared_prefix_burst": false,
    "thinking_mode": "default"
  },
  "source_recipe": null,
  "started_at": "2026-08-29T19:04:02Z",
  "tier": null,
  "timing_methodology": {
    "clock": "client time.perf_counter",
    "decode": "usage completion tokens after the first token divided by client-observed generation time; completion tokens may include reasoning tokens",
    "effective_prefill": "usage.prompt_tokens divided by client-observed time to first output; includes queueing, scheduling, prefill, and first-token work",
    "generation": "client-observed E2E minus time to first output",
    "mean_inter_token_latency": "client-observed generation time divided by completion tokens after the first token; not a raw per-token timestamp trace",
    "time_to_first_output": "request start through the first non-empty streamed reasoning or content delta",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "wall_clock_ms": 21223.415399999794
}
